Meta AI Confused Deputy Flaw Enables High-Profile Instagram Account Takeovers

Attackers hijacked multiple high-profile Instagram accounts by exploiting a confused deputy weakness in Meta’s AI-powered customer support chatbot. The attack was straightforward: the adversary asked the AI to link the target account to a new email address, successfully impersonating the legitimate owner without meaningful verification. Once linked, the attacker locked out the real owner. The flaw exposes a fundamental trust problem in AI-mediated account management — the model executed account-modification requests without sufficient identity validation. Meta has not confirmed a patch at time of writing; account owners should verify recovery options are set to personal, attacker-inaccessible contacts.